Effect of reconstruction parameters on cone beam CT trabecular bone microstructure quantification in sheep

Assessment of trabecular bone microstructure from CBCT images prior to implant surgery is a growing treatment planning protocol to ensure successful placement and clinical longevity of the dental implants [1]. This is because the trabecular bone microstructure has been shown to have major effects on primary stability during implant placement [2, 3]. In addition, primary stability can be accurately predicted based on the status of the trabecular bone microarchitecture prior to implant placement [4]. Nevertheless, the assessment of trabecular parameters from CBCT images may be compromised by the reconstruction parameters that are used for image segmentation, such as the reconstruction voxel size and threshold value.

Reconstruction voxel size is the voxel size that is used to reconstruct an image of raw data. A larger image reconstruction voxel size than the actual image acquisition voxel size is usually used to reduce the reconstruction time and to reduce computational expenses [5]. This is used in some practices such as analysing large-scale finite element models using a micro- CT-based image [5‐7]. However, it is presumed that the image accuracy will be decreased when using a larger reconstruction voxel size than the acquisition voxel size [8]. The effect of CBCT reconstruction voxel size has focused on image quality [9] and 3D measurement of dental morphology [10]. However, the effect on trabecular bone microstructure measurements remains unknown.

Segmentation is a process of separating images to either bone or other soft tissue structures which can be affected by the threshold values [11]. Many studies have used μCT to investigate the effect of threshold values on trabecular microstructure parameters such as trabecular number (Tb.N), thickness (Tb.Th), spacing (Tb.Sp) and bone volume fraction (BV/TV) of small animal models [12]. Although studies on scanning parameters are abundant in large animal models, the effect of reconstruction parameters using CBCT are scarce. Hence, as it has been suggested that CBCT can be used for trabecular bone measurement to aid surgical treatment planning, the CBCT reconstruction parameters that affect these measurements should be investigated.

Threshold values are commonly used for segmentation as it is the most straightforward and feasible segmentation method specifically when measuring bone volume [13]. Therefore, its effect on trabecular quantification is also investigated in this study. The findings of this study will provide clinicians with an evidence-based guideline in deciding the suitable reconstruction voxel size and threshold values to be used for their CBCT image reconstruction prior to trabecular bone microstructure analysis.

Five sectioned maxillae of adult Dorper male sheep (age ranging from 20 to 48 months and body weight 50–60 kg) were obtained from the Animal Experimental Unit, Faculty of Veterinary Medicine, University Putra Malaysia (UPM), Malaysia following approval from the Institutional Animal Care and Use Committee (IACUC), UPM (No. UPM/IACUC/AUP-R031). The samples were imaged using a CBCT system (Carestream 9000, Kodak) at the Division of Oral Radiology, Faculty of Dentistry, University of Malaya, Kuala Lumpur, Malaysia. The selected scanning parameters for CBCT were 76 μm³ for the voxel size, 4 × 4 cm FOV and 360° arm rotation. Images were acquired at 65kv, 6 mA and 10.8 s. The samples were fitted in a cylinder-shaped plastic container to reduce any possible movement during scanning. The images were subsequently converted into BMP and imported into DataViewer (v 1.5, SkyScan, Kontich, Belgium) to obtain the sagittal view for image processing and analyses in CTAnalyser software (v 1.15, SkyScan, Kontich, Belgium).

The assessments were performed twice by two examiners with a minimum interval of one week between the two measurements. Intraclass correlation coefficient (ICC) was employed to examine the intra-observer reliability and inter-rater agreement in reproducing the measurements. The trabecular bone microstructure parameters that were assessed include trabecular number (Tb.N), thickness (Tb.Th), separation (Tb.Sp) and bone volume fraction (BV/TV). One-way ANOVA was used to assess the differences obtained from three reconstruction voxel sizes and threshold values. Additionally, the Bonferroni test was used as a post hoc test to examine the significant differences between threshold values.

The enhanced resolution of cone-beam computed tomography (CBCT) images has significantly improved the measurement of trabecular bone microstructure [14, 15]. Unlike scanning parameters (voxel size, field of view and scanning rotation) [16], the effect of CBCT reconstruction parameters (reconstruction voxel size and threshold value) on diagnostic accuracy of trabecular bone microstructure has not been reported [17, 18].

The accuracy of micro-CT images in measuring trabecular bone microstructure can be compromised by the reconstruction voxel size [8, 19, 20]. Previous micro-CT studies have reported that the image quality cannot be improved by using a smaller reconstruction voxel size after originally scanning the sample using a large voxel size [8, 21]. This is due to an increase in image noise (9). Therefore, in the current study, CBCT images that were scanned using 76 μm³ were reconstructed using larger voxel sizes (100 and 200 μm³). The findings indicated no differences in trabecular bone measurements between different reconstructed images. However, our results differ to those reported in previously studies [8, 21] as the scanning voxel size (76 μm³) is almost 4 times larger than was used in micro-CT studies (21 μm³). It was demonstrated in micro-CT studies that only certain trabecular bone parameters namely BV/TV and Tb.Th are compromised by scanning parameters when reconstructing the images of 21 μm³ using larger voxel sizes (50 and 110 μm³) [8]. This is due to the fact that the trabecular bone parameters are significantly affected by the scanning voxel size rather than the reconstruction voxel size. However, this effect can only be observed when the difference between the scan and reconstruction voxel is very large [21]. The results of the present study showed that trabecular bone microstructure measurements are not influenced by CBCT reconstruction voxel size, although it is important to note other CBCT systems with different scanning and reconstruction parameters may generate different results.

Threshold values may influence the analysis of trabecular microstructure parameters in micro-CT images [22‐24]. Similarly, our study of CBCT images showed that a reducing the threshold value by 15% had a significant effect on all trabecular bone microstructure parameters (Tb.N = 0.03, Tb.Sp = 0.04, Tb.Th = 0.01, BV/TV = 0.00). However, increasing the threshold value from the global value had no significant (p < 0.05) effect on trabecular bone measurements (Table 4). The deviation of trabecular bone measurements might be due to partial volume effects that might alter the layers of voxels from the trabecular surface [25]. However, Tb.N measurements are insensitive to threshold variation if the selected threshold values are within the range of realistic values.

However, some limitations should be noted. In this study, the samples were not assessed using micro CT images due to computational constrains. Although strong correlations between micro CT and CBCT trabecular bone measurements have been largely described [26‐28], the comparison in varying the reconstruction parameters is highly recommended to further validate the accuracy of the current findings.

This study showed that trabecular bone microstructure measurements are not compromised by changing the CBCT reconstruction voxel size. However, measurements can be affected when applying a threshold value less than 15% of the recommended global value.